Applied Acoustics 73 (2012) 1211–1219

Contents lists available at SciVerse ScienceDirect

Applied Acoustics
journal homepage: www.elsevier.com/locate/apacoust

A review of parametric acoustic array in air

Woon-Seng Gan a,⇑, Jun Yang b, Tomoo Kamakura c
a
School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798, Singapore
b
The State Key Laboratory of Acoustics and the Key Laboratory of Noise and Vibration Research, Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190, China
c
Department of Communication Engineering and Informatics, The University of Electro-Communications, 1-5-1 Chofugaoka, Chofu-shi, Tokyo 182-8585, Japan

a r t i c l e i n f o a b s t r a c t

Article history: In this review paper, we examine some of the recent advances in the parametric acoustic array (PAA)
Available online 15 May 2012 since it was first applied in air in 1983 by Yoneyama. These advances include numerical modelling for
nonlinear acoustics, theoretical analysis and experimentation, signal processing techniques, implementa-
Keywords: tion issues, applications of the parametric acoustic array, and some safety concerns in using the PAA in
Parametric acoustic array air. We also give a glimpse on some of the new work on the PAA and its new applications. This review
Parametric loudspeaker paper gives a tutorial overview on some of the foundation work in the PAA, and serves as a prelude to
Directional sound beam
the recent works that are reported by different research groups in this special issue.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction 2. Theoretical framework of the parametric acoustic array

The discovery of the parametric acoustic array (PAA) has come a
long way (almost half century ago) since it was first theoretically
analyzed by Westervelt [1] in 1963. It has since moved from theory
and experimentation to implementation and application. Despite
several useful characteristics of the PAA, such as high directivity,
small size, and very small sidelobes, we are only just beginning to
witness some interesting innovations from commercial companies
in deploying the PAA for audio and speech applications. However,
there are still some technical challenges concerning ultrasonic
emitter performance, conversion efficiency, power consumption,
acoustic modelling and measurement, and digital signal processing
that are related to the PAA in air. This review paper provides a quick
overview of some of the important milestones achieved in the field
of the PAA, with the emphasis on creating the PAA in air. It also
serves as a preamble to some of the latest works that are reported
in this special issue.
This paper is organized as follows. A theoretical framework of the
PAA is first introduced in the next section. This is followed by a brief
description of the theory of audio parametric loudspeakers in
Section 3, and Section 4 outlines several interesting developments
of signal processing and modulation techniques for parametric
loudspeakers. Section 5 highlights some of the current state-of-
the-art implementation platforms used in parametric loudspeakers.
The issue of safety in parametric loudspeakers is also highlighted in
Section 6. Finally, Section 7 concludes this review paper.
⇑ Corresponding author.
E-mail addresses: ewsgan@ntu.edu.sg (W.-S. Gan), jyang@mail.ioa.ac.cn
(J. Yang), kamakura@ee.uec.ac.jp (T. Kamakura).
When two sinusoidal beams are radiated from an intense ultra-
sound source, a spectral component at the difference frequency is
secondarily generated along the beams due to the nonlinear inter-
action of the two primary waves. At the same time, spectral com-
ponents such as a sum-frequency component and harmonics are
generated. However, only the difference-frequency component
can travel an appreciable distance because sound absorption is
generally increased with frequency, and amplitudes of higher-fre-
quency components decay greatly compared with the difference
frequency. The secondary source column of the difference fre-
quency (secondary beam) is virtually created in the primary beam
and is distributed along a narrow beam, similar to an end-fire array
reported in antenna theory [1], as shown in Fig. 1. Consequently,
the directivity of the difference-frequency wave becomes very nar-
row. This generation model of the difference frequency is referred
to as the PAA.
The basic idea of the parametric array was originally conceived
by Westervelt about 50 years ago based on the scattering of sound
by sound [1]. When two primary waves of frequencies f1 and f2
(f2 > f1) are fully confined beams, he found that the angle, at which
the sound intensity of the difference frequency f = f2  f1 is reduced
by one-half (3 dB), is given approximately by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
hh  2aT =k; ð1Þ
where k is the wavenumber of the difference-frequency wave, and
aT is the total sound absorption coefficient of the primary waves.
When f1  f2, aT  2a1, where a1 is the absorption coefficient of
the primary wave of frequency f1. Eq. (1) illustrates how narrowing

0003-682X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.apacoust.2012.04.001
1212 W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219
Fig. 1. Generation of audible sound beam through the PAA.
Fig. 2. Axial sound pressure curves for the primary and secondary waves (theory).
The two primary waves of 38 kHz and 40 kHz from a sound source of 10 cm in
radius produce a 2-kHz difference frequency wave in the air. The initial source
pressures of the primaries are set to 125 dB. The dash line denotes an extrapolated
6 dB/dd line.
the secondary beam is realized by decreasing the primary frequen-
cies and/or by increasing the secondary frequency. Interestingly, the
directivity of a parametric source is thus independent of the source
aperture. This may be contrasted with linear theory, which states
the directivity of a sound source is dependent on two parameters,
namely driving frequency and source aperture. Strictly, however,
the primary beam is dependent on the source aperture of primary
waves, especially when the wavelength of the difference wave is
comparable with or smaller than the source aperture [2,3]. As an
example, let the primary frequency be around 40 kHz and the dif-
ference frequency be 2 kHz. The absorption coefficient of airborne
ultrasound is approximately 0.15 Neper/m at 40 kHz in ordinary
room conditions. Hence from Eq. (1), hh is predicted as 7.4°. In con-
trast, a large aperture source of at least 68 cm in radius is needed by
linear theory to realize such a narrow audio beam at 2 kHz.
As described above, the most remarkable acoustic property of
the parametric array is its sharp directivity at low frequencies.
Additionally, side-lobes, which usually exist for a directive sound
source, are suppressed considerably. Furthermore, small changes
Fig. 3. Sound pressure distributions of a parametric array (right) and an ordinary sound source (left). Both the sources have the same apertures of 10 cm in radius, radiating
sounds at a frequency of 2 kHz. The primary frequencies are 40 and 42 kHz.
in the high primary frequencies produce large relative changes in
the low secondary frequency. In other words, the parametric sound
is equivalent to a low-Q source even if high-Q ultrasound sources
are used.
In order to accurately evaluate parametrically generated sound
fields, diffraction as well as nonlinearity in primary waves are ta-
ken into account. The most useful and traditional model equation
for such field evaluation is the Khokhlov–Zabolotskaya–Kuznetsov
(KZK) equation, which combines nonlinearity, dissipation, and dif-
fraction of a directive sound beam in the same order of magnitudes
[4]. This model equation is described as:
@2p c0 2 d @3p b @ 2 p2
0 ¼ r ? p þ 3 03 þ ; ð2Þ
@z@t 2 2c0 @t 2q0 c30 @t 02
where p is the sound pressure, c0 is the sound speed, q0 is the med-
ium density, d is the sound diffusivity that is related to sound
absorption, and b is the nonlinearity coefficient, which is equal to
1.2 for air. Moreover, r2? ¼ @ 2 =@x2 þ @ 2 =@y2 is a Laplacian that oper-
ates in the x–y plane perpendicular to the axis of the beam (z axis),
and t 0 ¼ t  z=c0 is the retarded time. It is quite cumbersome to
solve analytically the KZK equation even when nonlinearity is weak
[3]. Especially, when nonlinearity is moderate or strong, we resort
to numerical computation methods such as a finite difference
scheme to obtain the solution. Since the KZK equation is derived un-
der the parabolic approximation, we have to pay attention to its
applicability; the upper limit of beam angle is restricted to the par-
axial region that is within about 20° from the z axis [5].
The fundamental characteristics of parametric sound in air are
demonstrated in Figs. 2 and 3. An ultrasound source, whose circu-
lar aperture is 10 cm in radius, radiates bifrequency waves of
38 kHz and 40 kHz with the same pressure amplitudes of
p0 ¼ 50 Pa (125 dB re. 20 lPa) at the source surface. Fig. 2 shows
the axial pressure profiles of the primary waves and the parametric
sound of the difference frequency at 2 kHz [6]. Note that the ampli-
tude of the parametric sound increases with propagation, attains
the maximum at about 1.5 m from the source, and then decreases
gradually. Unfortunately, the pressure level of the difference-fre-
quency wave is generally 40 dB or more lower than the level of
the primary waves, except in the farfield. Additionally, unwanted
harmonic sounds, such as a 4-kHz component, are prominently
generated in a field that is a few metres away from the source.
Hence, it is of importance to reduce such harmonic distortions
and cross-modulation distortions as much as possible in designing
parametric loudspeakers.
Fig. 3 shows the comparison of pressure distributions produced
by a parametric array and an ordinary piston source under the
conditions that the source radii are both 10 cm and audible
 W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219
frequencies of 2 kHz. It is stressed that the directivity of the para-
metric array is dramatically sharper than that of the ordinary
source.
3. Realization of parametric loudspeakers
Since the discovery of the PAA by Westervelt, more than several
hundred papers and technical reports on the topics have been pub-
lished so far from theoretical and experimental points of view.
However, almost all the papers before 1975 were focused on
exploratory examinations and underwater applications [7,8]. In
1965, Berktay provided a simple expression on the pulse produced
by the self-demodulation of a pulsed carrier [9], serving as the ba-
sis of predicting the far-field array response of a parametric loud-
speaker. The expression states that the modulated signal (or
audio difference frequency) pressure p(t0 ) along the axis of propa-
gation is proportional to the second time-derivative of the square
of the envelope of the amplitude-modulated ultrasound carrier as
follows:
2
bp20 a2 d 2 0
pðt0 Þ  E ðt Þ;
16q0 c40 za0 dt 02
where p0 is the pressure source amplitude, a is the source radius, a0
is the absorption coefficient of the ultrasound carrier, and E(t0 ) is the
modulation envelope function of the carrier. Eq. (3) shows that the
demodulated signal is principally proportional to the size of the
ultrasound source, the pressure amplitude of primary wave, and
the envelope function form. However, the high-frequency carrier
reduces the amplitudes of parametric sounds due to the increase
of carrier sound absorption. Therefore, higher audible sound pres-
sures at a distance can be achieved by changing the values of the
above four parameters. The application of Berktay’s model is limited
to cases where the primary source pressure is relatively low so that
the parametric array is determined by small-signal absorptions of
the primary waves. Later, Merklinger advanced Berktay’s analysis
to the case where the ultrasound carrier is sufficiently intense to
cause nonlinear attenuation [10]. His expression is stated as
 
0 
Sp0 @2 1 bxp0 Eðt Þ
pðt0 Þ ¼ Eðt 0
Þ tan ; ð4Þ
4xpc0 z @t 0 2
4aq0 c30
where S is the aperture area of an ultrasound source and is given as
pa2 for a circular aperture with radius a. In Eq. (4), the solution is
approximately subjected to p0  4aq0 c30 =bx as
@2
pðt0 Þ / p20 2
E2 ðt 0 Þ:
@t 0
This solution indicates that the amplitude of produced paramet-
ric sound is proportional to the square of the envelope and is the
same form as Berktay’s expression. Whereas, when nonlinearity
is strong and the condition p0  4aq0 c30 =bx is satisfied, the
expression becomes
@2
pðt0 Þ / p0 02
jEðt0 Þj:
@t
We note that the parametric sound is proportional to the ampli-
tude of the carrier itself. It is therefore necessary to appropriately
take account of the demodulation processes being changed from
Eq. (5) to Eq. (6) in response to the primary wave amplitude when
designing suitable parametric loudspeakers [11]. The following
section presents the signal processing and modulation techniques
that are derived from the Berktay’s model (Eq. (3) or Eq. (5)).
1213
4. Signal processing and modulation techniques
When employing the PAA principle for directional sound, the
chosen primary wave usually lies beyond the human hearing
range, typically at around 40 kHz, which is amplitude-modulated
by audio signals. Thus, amplitude-modulated ultrasound wave
has a carrier, upper and lower side-band components, resulting
in reproduction of the audible sound in air due to the nonlinear
interaction of the carrier and each side-band in the ultrasound
beams. Needless to say, the directivity of the reproduced audible
sound is very sharp owing to the characteristic of the parametric
array. It shall be shown that the sound pressure level and harmonic
distortion of the demodulated signal are proportional to the mod-
ulation index, therefore care must be exercised to determine the
modulation index used in amplitude modulation (AM). We will
discuss different types of modulation techniques that can reduce
the distortion introduced by the self-demodulation process in the
following subsections. A single-tone analysis is used to evaluate
the performance of reducing distortion for different pre-processing
and modulation techniques.
ð3Þ 4.1. Double sideband amplitude modulation
In 1983, Yoneyama et al. [12] proposed a parametric loud-
speaker system, which used the conventional AM or the double
sideband amplitude modulation (DSBAM). The modulation enve-
lope for their parametric loudspeaker system is given as
E(t0 ) = 1 + mg(t0 ), where m is the modulation index and g(t0 ) is the
input signal. The block diagram of the DSBAM is shown in Fig. 4,
where sin(x0t0 ) is the ultrasonic carrier.
However, this original PAA system would cause high total har-
monic distortion (THD). This phenomenon was validated by their
experiments, where the second harmonic was of similar level to
the fundamental signal in the case of a single-tone input for high
m. Fig. 5 summarizes the THD vs. modulation index m for DSBAM.
This figure shows that DSBAM is not a preferred technique because
it incurs high distortion at high m. Moreover, a high modulation in-
dex is required to produce a demodulated signal with desirable
high sound pressure level at the expense of increasing distortion.
By reducing the modulation index, there is a tradeoff between
sound pressure level of the demodulated signal and lower distor-
tion, which is not desirable for practical applications. Therefore,
DSBAM is seldom used as the modulation technique for parametric
loudspeakers, except for the case where DSBAM is employed to
evaluate the acoustic performances of the PAA. In the following
sections, several modified amplitude modulation techniques,
which achieve high demodulated sound pressure level with re-
duced distortion, will be presented.
ð5Þ
4.2. Square-root amplitude modulation
Based on the Berktay’s model expressed in Eq. (3), many other
attempts [13–22] have been implemented to improve the quality
of the demodulated signal. In 1984, Kamakura et al. [13,14] pre-
sented a square-root AM (SRAM) method that applied an envelope
function:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð6Þ Eðt 0 Þ ¼ 1 þ mgðt 0 Þ: ð7Þ
g (t′) m
sin ωct ′
Fig. 4. Block diagram of the double sideband amplitude modulation.
1214
Total Harmonic Distortion (THD) vs. Modulation Index
80
70
Total Harmonic Distortion (%)
60
50
40
30
20
10
0 parameters were presented for directional speech reproduction.
0 0.1 0.2 0.3 0.4 0.5
Modulation Index, m
Fig. 5. Total harmonic distortion vs. modulation index.
0.5
Fig. 6. Block diagram of square-root amplitude modulation.
The block diagram of the SRAM method is demonstrated in
Fig. 6. Compared to the conventional DSBAM method, lower THD
values have been achieved. However, the ultrasonic emitter with
large bandwidth is required to generate the infinite harmonics
introduced by the square-root operation. The same technique
was also adopted by Pompei [15] and Kite et al. [16].
4.3. Single-sideband amplitude modulation
Another modulation method was proposed to reduce both the
distortion and power consumption in driving parametric loud-
speaker based on the single-sideband amplitude modulation
(SSB-AM) [17], as shown in Fig. 7. The major advantage of the
SSB-AM method is that it produces a similar envelope as that in
the SRAM method, with only half the bandwidth of the SRAM
method. In the case of two primary waves, there is no difference
between the envelope produced by the SSB-AM and the SRAM
methods, as shown in Fig. 8. This feature in SSB-AM implies that
it is necessary to use a high bandwidth emitter, as in the case of
SRAM, to achieve low THD performance. However, in case where
multiple primary waves or a broadband signal (such as speech)
modulates a single primary carrier, envelope error occurs.
Hence, Croft et al. [18,19] proposed a recursive SSB-AM (RSSB-
AM) method to approximate the envelope generated by the SRAM
method, as shown in Fig. 9d. The RSSB-AM method consists of the
SSB modulator and the nonlinear demodulator (NLD), as shown in
Fig. 9a and b, respectively. The role of the NLD is to calculate the
square of the envelope, which models the nonlinear acoustic prop-
agation in air by assuming that the second-time derivative effect in
the Berktay’s model can be perfectly compensated. These two
blocks are combined into a distortion model (DM) in Fig. 9c and
subtracted from the original input g(t) = g0(t) to obtain the distor-
tion di(t) at the output of the ith stage, where i = 1, 2, . . . , q. The dis-
tortion is progressively reduced by each DM stage, and higher
reduction is achieved by cascading several DM stages. Because of
the high complexity of the RSSB-AM method, a high-speed proces-
sor must be employed to achieve real-time performance. A detailed
W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219
Fig. 7. Block diagram for single sideband modulation with carrier.
investigation on the analytical performance of the RSSB-AM meth-
od was given in [20], where an RSSB-AM method with optimal
0.6 0.7 0.8 0.9 1
4.4. Modified amplitude modulation
A different modulation technique known as the modified AM
(MAM) method, which is a class of hybrid AM and SRAM methods
based on the orthogonal amplitude modulation, is proposed in
[21,22]. The block diagram is shown in Fig. 10. It has the flexibility
to scale the relative bandwidth requirement to match the band-
width of the ultrasonic emitters, and also provides complexity
scaling.
5. Implementation Issues
In this section, we look into some of the implementation issues
on developing the parametric loudspeakers in analog and digital
circuits. The parametric loudspeaker system consists of three main
parts, namely, signal processing, amplifier, and ultrasonic emitter,
as shown in Fig. 11. Some of the early parametric loudspeakers
used analog circuit elements to perform the signal processing
block. Fig. 12 exemplifies a typical analog circuit for implementing
the PAA system [23]. A peak level detector, which is fed by the
input signal, is employed to prevent overmodulation and ensure
the envelop signal generated from the adder is always positive.
The peak level detector can be designed with a voltage comparator
(LM111), and the adder is simply achieved by using an operational
amplifier. Next, the envelope signal is pre-processed by applying a
square-root operation using a four-quadrant analog multiplier
(AD734). This pre-processing step has been previously explained
in Section 4.2 to reduce the distortion produced during the nonlin-
ear interaction of the air propagation. The carrier signal is gener-
ated using a high-frequency, precision waveform generator
(MAX038). In the final step, the carrier signal is multiplied with
the pre-processed signal to form the modulated signal. The analog
development of a parametric loudspeaker has the advantages of
low cost and simple structure. However, the analog circuit is less
flexible, less stable, and it is inefficient for implementing compli-
cated functions and operations using analog components, such as
a Hilbert filter in the SSBAM scheme. These disadvantages have
limited the development of high performance parametric
loudspeaker.
A digital signal processor has the ability to process more com-
plicated operations without incurring extra cost, size, and power
consumption compared to the analog circuit system. A field pro-
grammable gate array (FPGA) is an attractive digital processing
platform to implement the parametric loudspeaker due to its flex-
ible configuration and high performance. FPGA has a programma-
ble feature of logic cells, which can be interconnected to perform
signal processing algorithms either in sequential or parallel opera-
tion. Thus, the FPGA can be configured for high-speed parallel pro-
cessing with the tradeoff of more logic gates used.
 W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219 1215

1 1
0.8 0.8
0.6 0.6
0.4 0.4
Amplitude

Amplitude
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 -1
2 4 6 8 10 12 14 16 18 2 4 6 8 10 12 14 16 18
-4 -4
time (sec) x 10 time (sec) x 10

(a) (b)
220 220
200 200
180 180
160 160
Amplitude (dB)

Amplitude (dB)
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
2 2.5 3 3.5 4 4.5 5 5.5 6 2 2.5 3 3.5 4 4.5 5 5.5 6
4 4
frequency (Hz) x 10 frequency (Hz) x 10

(c) (d)
Fig. 8. Time domain plots of modulation envelopes of (a) SRAM, and (b) SSB; frequency spectra of the modulated signal of (c) SRAM and (d) SSB.

g (t′) m

2
sin ωct ′

(a) cos ωct ′

0.5

Fig. 10. Block diagram of the MAM method.

(b)

(c) Fig. 11. Block diagram of parametric loudspeaker.

Fig. 13 shows the generic block diagram for implementing dig-

(d)
Fig. 9. Block diagram of the RSSB-AM and its components. (a) SSB modulator, (b)
NLD, (c) single DM stage, and (d) cascade of DM stages to form the RSSB-AM.
ital pre-processing and modulation in the parametric loudspeaker
[24]. A dynamic range control processor consists of a compressor
and a limiter. By compressing the dynamic range of the input
speech or musical signal with a constant gain, the reproduced
sound can be perceived clearly without noticeable distortion. The
limiter is used to prevent the input signal from clipping, which
1216 W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219

+15V

LM111
Input

Envelope signal
OP27
-15V

Peak level detector

Carrier generator Multiplier

Frequency
Modulated signal
REF Z1 W X1 W

DADJ AD734 AD734

MAX038
Z2 Y1
IIN
FADJ

OUT

Fig. 12. Analog implementation of the PAA system.

Fig. 13. Digital processing of parametric loudspeaker.

Analog
FPGA Board Input Interface Board

Analog
Data I/O Output
Clock
Signals

Switch
Signals

LEDs

Reset

+ 3.3 V
GND
& +5V

Fig. 14. Interconnection between FPGA board and interface board.

can cause perceivable distortion in the audible output. The output
of the dynamic range control processor is passed to the pre-distor-
tion processing block. This block pre-processes the input signal to
reduce the amount of distortion produced from the higher-order
nonlinear interaction under nonlinear acoustics propagation. The
pre-processing methods have been previously discussed in Section
4.2. Next, an automatic carrier level controller performs automatic
gain control to the carrier signal based on the input level. The gain
is calculated to proportionally scale the oscillator output with
respect to the input level, which can result in significant reduction
in power consumption. The sinusoidal oscillator, which can be
implemented using either a recursive algorithm or a look-up table,
is used to generate the required (single or orthogonal) carrier sig-
nal(s). The final process modulates the dynamic carrier signal using
the pre-processed audible signal.
Fig. 14 shows the connection between the FPGA chip and the
audio analog–digital convertor (ADC) and the ultrasonic-frequency
digital-to-analog convertor (DAC). The FPGA chip performs the dig-
ital signal processing, as described in Fig. 13. An audio codec, which
consists of a two-channel 16-bit 96 kHz ADC and 192 kHz DAC, is
 W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219
Fig. 15. Block diagram of Class-D amplifier for the PAA.
used in our digital implementation. This audio codec is suitable for
emitting ultrasonic frequencies in the range of 30–50 kHz, which
falls within the resonating frequency of most ultrasonic emitters
used in parametric loudspeakers. A low-pass filter on the interface
board performs signal conditioning and the voltage regulator pro-
vides a clean and stable DC supply across the board.
In order to create a parametric array effect in air, the modulated
signal must be amplified before being fed into the ultrasound emit-
ter. The Class-D power amplifier is widely used in the parametric
loudspeaker due to its efficiency and small size. Fig. 15 shows
the block diagram of the Class-D amplifier used in the parametric
loudspeakers. In the preceding FPGA implementation, the digital
modulated signal can be directly converted to a pulse train using
the pulse modulator block, without the need of a DAC. Two pulse
modulation techniques are commonly used: pulse width modula-
tion (PWM) and pulse density modulation (PDM). PWM generates
pulses with widths proportional to the input amplitude, while
PDM generates pulses with fixed width but with the density of
the pulse train dependent upon the amplitude [25]. PWM is more
efficient than PDM, but generates higher harmonic distortion. In
the Class-D output stage, the incoming pulses are amplified so as
to drive the ultrasound emitter. Since the switching between the
on and off stages is not ideal, the output stage will add distortion
and noise to the signal channel. This problem can be significantly
alleviated if an error correction technique [26] is adopted. Finally,
the analog signal is derived by using a demodulation filter, which
should be carefully designed to remove the high frequency prod-
ucts and maximize the power transfer.
6. Applications of the parametric acoustic array
The PAA has commonly been used in display kiosk, entertain-
ment, communication, and personal messaging systems [27,28]
due to its ability to create a highly directional sound beam using
a small-surfaced emitter. In this section, we highlight some recent
applications of the PAA that show promising potential in exploiting
the unique features of the PAA.
Traditionally, in active noise control (ANC) systems, omni-direc-
tional loudspeakers are utilized as control sources to suppress the
noise in the targeted area. However, the sound pressure level (SPL)
of the noise at locations away from the target area may actually in-
crease. To solve this problem, Brooks et al. [29] investigated the
feasibility of using a parametric loudspeaker as a control source
in ANC systems. Experimental results showed that an ANC system
using a parametric loudspeaker could reduce the SPLs at the con-
trol points without increasing noise level in other areas. Kidner
et al. [30] proposed a method of combining the PAA and virtual
sensing techniques to create localized zones of quiet. Tanaka and
Tanaka [31,32] presented a novel idea of using a steerable PAA
based upon phased array theory with an optimal control law to
track a moving target point without rotating the parametric loud-
speaker mechanically. Furthermore, the same research group [33]
proposed a novel method to use the reflected sound wave, which
is produced by the parametric loudspeaker, on the target source
1217
to collocate both the primary source and the control source. There-
fore, global noise control could be achieved because the distance
between the two sources became theoretically nullified. In another
work, Komatsuzaki and Iwata [34] compared the interfered sound
field resulting from using the parametric loudspeaker and omni-
directional loudspeaker as control sources to achieve local noise
control. Three different constructions of the primary source and
the control source were considered: opposed placement, orthogo-
nal placement, and coaxial placement. Results showed that the
parametric loudspeaker can be used as a control source to mitigate
sound locally without increasing noise level in other areas.
In the area of personal communication, Nakashima et al. [35,36]
mounted two parametric loudspeakers on a prototype mobile
phone to deliver a private sound field (or personal sound) to the
users. Each parametric loudspeaker consisted of 16 piezoelectric
emitters. The SPL of the audible sound on the central axis was mea-
sured to be more than 70 dB at a distance of about 50 cm away
from the phone. Because of the directivity of the PAA, the sound
pressure difference between the two ears is approximately 15 dB,
which is helpful in binaural sound reproduction.
Due to the high directivity of sound wave at low frequencies,
the PAA has also been found to be suitable in detecting concealed
objects, such as land mines and weapons. Haupt and Rolt [37] used
the PAA to excite the buried mines from a safe distance. The vibra-
tion signatures of the mines were measured by the laser vibrome-
ter to locate the position of the mines. In another work, Achanta et
al. [38] presented a novel method to use the PAA to detect con-
cealed weapons. The metallic and non-metallic materials under
clothing, as well as abnormality in handheld devices, can be
scanned and detected by the directional sound at low frequencies
generated by the PAA system.
The PAA can also be deployed to measure the acoustic perfor-
mance of materials. The application of the PAA in this field was
introduced by Humphrey [39–41]. His group used a PAA as a sound
source to determine the reflection loss and transmission loss of
panels in a vessel which simulated the ocean environment. With
the highly directional sound beams, the effects of diffraction from
the panel edges were reduced. Castagnede’s group used a PAA and
a microphone to measure the absorption coefficient of material at
normal incidence in air under the excitation of impulse signals
[42,43]. Kuang et al. introduced the transfer function method into
the process [44], which allowed measurements to be taken in situ.
Moreover, this method can be combined with the phase-cancella-
tion method proposed by Kamakura [45] to effectively remove spu-
rious noise for accurate measurements [46].
Many theoretical and/or experimental data have also been pre-
viously reported for the PAA in underwater applications. The
acoustic features of the parametric array in water are basically
the same as that in air, with sharp directivity at low frequency
and with suppressed side-lobes. The technique that uses paramet-
ric sound propagation, such as navigation, communication with
other vessels, or detection of targets is sometimes referred to as
the parametric sonar. The parametric sonar is one of the most ma-
tured technologies developed and evaluated in underwater acous-
tics, and its academic papers and technical reports are summarized
in scientific monographs and reference materials [7,8]. Applica-
tions and challenges of PPA in shallow water sonar and sub-bottom
profilers are also current research of interest.
7. Safety issues
In this section, we shall discuss how airborne ultrasound affects
human body. Almost 99.9% of the ultrasound energy is reflected by
the human skin due to the high impedance mismatch between air
and the skin. Generally, the major effects of ultrasound exposure in
1218 W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219

Table 1
Guidelines for the safe usage of ultrasound that are recommended in various countries. All values in decibels are upper limits for whole-day exposure. Table extracted from [47].

One-third-octave band centre frequency (kHz) 10 12.5 16 20 25 31.5 40 50 63 80 100

Source Band level (dB)
IRPA, 1984 – – – 75 110 110 110 110 110 110 110
Australia, 1981 75 75 75 110 110 110 110 110 – – –
USA, 1981 80 80 80 105 110 115 115 115 – – –
Canada, 1980 80 80 80 110 110 110 110 110 – – –
Sweden, 1978 – – – 105 110 115 115 115 115 115 115
USSR, 1975 – 75 85 110 110 110 110 110 110 110 110
Norway, 1978 – – – – —120 (octave)— —120 (octave)— –

practice are induced through the ear. This is especially obvious if
the ultrasound level is extremely high and may cause unpleasant
sensation, headache, fatigue, and nausea [47]. Moreover, the symp-
toms vary from person to person. These effects are temporary and
are relieved by removing the ultrasound in most situations. It was
reported that long hours of exposure may cause threshold shift in
hearing loss [48]. However, there is no report that describes the
cause-and-effect relationship between hearing loss and exposure
of the ear to high frequency ultrasound.
For half a century, considerable efforts have been carried out to
find the human exposure limits of ultrasound as carried out by var-
ious countries and several health organizations. Such guidelines
have been presented for industrial workers using ultrasonic de-
vices, such as cleaning, drilling, and soldering from health and
safety protection points of view. Most of the guidelines recom-
mend an upper limit of ultrasound between 100 and 115 dB for fre-
quencies above 20 kHz as the admissible sound pressure level,
which will not cause audiometric hearing loss [47].
Ultrasonic devices obviously produce intense ultrasound,
whose frequencies are above 20 kHz. Additionally, these devices
generate wide band noisy audible by-products, which are possible
sources of annoyance for workers, rather than the inaudible ultra-
sound. Table 1 shows the maximum permissible sound pressure
levels [47] for the safe usage of ultrasound in various countries
and organizations. Basically, the pressure levels in the table are ap-
plied to continuous exposure for an 8-h working day. Some reports
indicate that the exposure limit levels may be increased for shorter
exposure durations, as shown in Table 2. For example, the Interna-
tional Radiation Protection Association (IRPA) committee recom-
mends that the permissible levels may be increased by 3 dB for a
duration of 2–4 h. However, the maximum limit prescribed by
the Health Canada is 110 dB above 25 kHz and it is independent
of duration [48,49].
In the parametric loudspeaker, the attained sound pressure
level can reach 120 dB, especially in the nearfield. Depending on
the dimension of parametric loudspeaker and its initial sound pres-
sure, a pressure level of 120 dB can be attained at 1 m from the
ultrasonic emitter, although such high pressure is limited to the
on-axis beam. Generally, the pressure level decreases beyond the
Rayleigh distance to less than 100 dB at 8 m. However, it is not
appropriate to readily apply the permissible pressure levels in Ta-
ble 2 as safety guidelines for operating the parametric loudspeak-
ers, since the ultrasound waves emitted from the parametric
loudspeakers consist of relatively narrow spectral bands around
the carrier frequency. This is fundamentally different from ultra-
sonic devices that radiate wideband ultrasound as well as broad-
band audible sound in space. However, as the specific response
to ultrasound frequencies is still unclear, the maximum permissi-
ble levels of 105–110 dB should be kept in mind even for paramet-
ric loudspeakers in short time operation because of questionable
problems in auditory and non-auditory biological effects of intense
ultrasounds around the carrier frequencies, such as 40 kHz.
Recently, Lee et al. [50] suggested from a physiological point of
view that the burden of a parametric loudspeaker is smaller than
Table 2
Supplement for exposure duration less than 8 h. Table extracted from [47].
Source Duration (h) Correction (dB)
IRPA, 1984 2–4 +3
1–2 +6
<1 +9
Sweden, 1978 1–4 +3
<1 +9
USSR, 1975 1–4 +6
1=
4 –1 +12
<1=4 +18
a conventional loudspeaker in listening tests of short-sentence
speech. Unfortunately, their report does not describe the relation-
ship between ultrasound pressure levels and physiological effects
of ultrasound on the auditory function of human being.
In summary, one of the important considerations in designing
and developing a parametric loudspeaker is to examine how to re-
duce the carrier ultrasound to a safe pressure level without com-
promising the performance of parametric sound.
8. Conclusions
This review paper serves as a quick tutorial reference to readers,
who are interested to further explore and extend this technology,
and bring this technology to other application areas. This review
outlined some of the important theoretical development of the
PAA, and gave an overview of different technologies, implementa-
tion techniques, applications and safety issue for parametric loud-
speakers. In particularly, we describe the current state-of-the-art
technology in pre-processing and modulating the audible signal
before sending to the ultrasonic emitters. However, these pre-pro-
cessing and modulation techniques are mainly based on the Berk-
tay’s model, which is only approximated solution for the sound
beam propagation in weak nonlinearity. It is the quest for future
work to obtain a more accurate and simple-to-use model that
can be used to derive a new pre-processing technique to equalize
the nonlinear distortion. Also, the computation cost for imple-
menting different pre-processing and beam control techniques
must be taken into consideration in building the next-generation
parametric loudspeakers, where new requirements, such as high
audio quality, electronic steerable, and accurate audio beam con-
trol are integrated into a personal sound entertainment system
and other novel applications, and some of which are being docu-
mented in this special issue.
Acknowledgements
The respective authors would like to acknowledge the financial
supports by the Singapore Ministry of Education Academic Re-
search Fund Tier 2, under research Grant MOE2010-T2-2-040, by
the National Natural Science Foundation of China under Grant
 W.-S. Gan et al. / Applied Acoustics 73 (2012) 1211–1219
No. 11004217 and Grant No. 11074279, and by a Grant-in-Aid for
Scientific Research (B) No. 22360094 from the Japan Society for the
Promotion of Science.
The authors would like to thank Mr. Ji Wei for his careful effort
in editing the drawings, cross-checking, and proof-reading this
manuscript.
References
[1] Westervelt PJ. Parametric acoustic array. J Acoust Soc Am 1963;35(4):535–7.
[2] Naze J, Tjøtta S. Nonlinear interaction of two sound beams. J Acoust Soc Am
1965;37(1):174–5.
[3] Hamilton MF. Sound beams. In: Hamilton MF, Blackstock DT, editors.
Nonlinear acoustics. New York: Academic Press; 1997 [chapter 8].
[4] Hamilton MF, Morfey CL. Model equations. In: Hamilton MF, Blackstock DT,
editors. Nonlinear acoustics. New York: Academic Press; 1997 [chapter 3].
[5] Jensen FB, Kuperman WA, Porter MB, Schmidt H. Computational ocean
acoustics. New York: Springer Verlag; 2000 [chapter 6].
[6] Kamakura T, Aoki K, Sakai S. A highly directional audio system using a
parametric array in air. In: Proc 9th Western Pacific acoustics conf, Seoul; June
26–28, 2006. p. 1–8.
[7] Novikov BK, Rudenko OV, Timoshenko VI. Nonlinear underwater
acoustics. New York: AIP Press; 1987.
[8] Moffett MB, Konrad WL. Nonlinear sources and receivers. In: Crocker MJ,
editor. Encyclopedia of acoustics. New York: John Wiley & Sons; 1997. p.
607–17.
[9] Berktay HO. Possible exploitation of nonlinear acoustics in underwater
transmitting applications. J Sound Vib 1965;2(4):435–61.
[10] Merklinger HM. Improved efficiency in the parametric transmitting array. J
Acoust Soc Am 1975;58(4):784–7.
[11] Aoki K, Kamakura T, Kumamoto Y. Parametric loudspeaker – characteristics of
acoustic field and suitable modulation of carrier ultrasound. Electron Commun
Jpn 1991;74(9):76–82.
[12] Yoneyama M, Fujimoto J, Kawamo Y, Sasabe S. The audio spotlight: an
application of nonlinear interaction of sound waves to a new type of
loudspeaker design. J Acoust Soc Am 1983;73(5):1532–6.
[13] Kamakura T, Yoneyama M, Ikegaya K. Development of a parametric
loudspeaker for practical use. In: 10th International symposium on
nonlinear acoustics, Kobe; 1984. p. 147–50.
[14] Kamakura T, Yoneyama M, Ikegaya K. A study for the realization of a
parametric loudspeaker. J Acoust Soc Jpn 1985(June):1–18.
[15] Pompei FJ. The use of airborne ultrasonics for generating audible sound beams.
J Audio Eng Soc 1999;47(9):726–31.
[16] Kite TD, Post JT, Hamilton MF. Parametric array in air: distortion reduction by
preprocessing. In: Proc 16th ICA and 135th meeting of the ASA, vol. 2; 1998. p.
1091–2.
[17] Sakai S, Kamakura T. Dynamic single sideband modulation for realizing
parametric loudspeaker. In: AIP conf proc; 2008. p. 613–6.
[18] Croft JJ, Spencer ME, Norris JO. Modulator processing for a parametric speaker
system, US 20080063214A1; 2003.
[19] Croft JJ, Norris JO. Theory, history, and the advancement of parametric
loudspeaker. White paper of American Technology Corporation; 2001.
[20] Ji P, Gan WS, Tan EL, Yang J. Performance analysis on recursive single-sideband
amplitude modulation for parametric loudspeakers. In: IEEE international
conference on multimedia and expo (ICME), Singapore; 2010. p. 748–56.
[21] Tan EL, Gan WS, Ji P, Yang J. Distortion analysis and reduction for the
parametric array. In: 124th Convention of the audio engineering society,
Amsterdam; 2008.
[22] Tan EL, Ji P, Gan WS. On preprocessing techniques for bandlimited parametric
loudspeakers. Appl Acoust 2010;71(5):486–92.
[23] Pompei J. Sound from ultrasound: the parametric array as an audible sound
source. PhD thesis. School of Architecture and Planning, Massachusetts
Institute of Technology; 2002.
[24] Karnapi FA, Gan WS, Chong YK. FPGA implementation of parametric
loudspeaker system. Microprocess Microsyst 2004;28(5–6):261–72.
1219
[25] Stark S. Direct digital pulse width modulation for class D amplifiers. Master
thesis. Department of Electrical Engineering, Linköping University, Sweden;
2007.
[26] Nielsen K. A review and comparison of pulse width modulation (PWM)
methods for analog and digital input switching power amplifiers. Bang &
Olufsen A/S, Denmark and Institute of Automation, Technical University of
Denmark.
[27] Pompei J. Audio spotlight. Sound Commun 2009;55(7):54–7.
[28] Gan WS, Tan EL, Kuo SM. Audio projection: directional sound and its
application in immersive communication. IEEE Signal Process Mag
2011;28(1):43–57.
[29] Brooks LA, Zander AC, Hansen CH. Investigation into the feasibility of using a
parametric array control source in an active noise control system. In: Proc
acoustics 2005, Busselton; 2005.
[30] Kidner MRF, Petersen C, Zander AC, Hansen CH. Feasibility study of localized
active noise control using an audio spotlight and virtual sensors. In: Proc
acoustics 2006, Christchurch; 2006.
[31] Tanaka N, Tanaka M. Active noise control using a steerable parametric array
loudspeaker. J Acoust Soc Am 2010;127(6):3526–37.
[32] Tanaka N, Tanaka M. Active noise control using a parametric array loudspeaker.
In: 17th International congress on sound and vibration, Cairo; 2010.
[33] Tanaka N, Tanaka M. Mathematically trivial control of sound using a
parametric beam focusing source. J Acoust Soc Am 2011;129(1):165–72.
[34] Komatsuzaki T, Iwata Y. Active noise control using high-directional parametric
loudspeaker. J Environ Eng 2011;6(1):140–9.
[35] Nakashima Y, Yoshimura T, Ohya T. Prototype of parametric array loudspeaker
on mobile phone and its acoustical characteristics. In: 118th Convention of the
audio engineering society, Barcelona; 2005.
[36] Nakashima Y, Yoshimura T, Naka N, Ohya T. Prototype of mobile super
directional loudspeaker. NTT DoCoMo Tech J 2006;8(1):25–32.
[37] Haupt RW, Rolt KD. Standoff acoustic laser technique to locate buried land
mines. Lincoln Lab J 2005;15(2):3–22.
[38] Achanta A, McKenna M, Guy S, Malyarenko E, Lynch T, Heyman J, et al.
Nonlinear acoustic concealed weapons detection. Mater Eval
2005;63(12):1195–202.
[39] Humphrey VF. The measurement of acoustic properties of limited size panels
by use of a parametric source. J Sound Vib 1985;98(1):67–81.
[40] Humphrey VF. The influence of the plane wave spectrum of a source on
measurements of the transmission coefficient of a panel. J Sound Vib
1986;108(2):261–72.
[41] Humphrey VF, Robinson SP, Smith JD, Martin MJ, Beamiss GA, Hayman G, et al.
Acoustic characterization of panel materials under simulated ocean conditions
using a parametric array source. J Acoust Soc Am 2008;124(2):803–14.
[42] Castagnede B, Saeid M, Moussatov A, Gusev V, Tournat V. Reflection and
transmission at normal incidence onto air-saturated porous materials and
direct measurements based on parametric demodulated ultrasonic waves.
Ultrasonics 2006;44(2):221–9.
[43] Castagnede B, Moussatov A, Lafarge D, Saeid M. Low frequency in situ
metrology of absorption and dispersion of sound absorbing porous materials
base on high power ultrasonic non-linearly demodulated waves. Appl Acoust
2008;69:634–48.
[44] Kuang Z, Wu M, Ye C, Yang J. Method for measuring the absorption coefficient
of sound absorbing materials in situ. Acta Acoust 2010;35(2):162–8 [in
Chinese].
[45] Kamakura T, Sakai S, Nomura H, Akiyama M. Parametric audible sounds by
phase-cancellation excitation of primary waves. In: Acoustics08, Paris; 2008.
[46] Kuang Z. Investigation on the measurement of absorption coefficients in situ
with the parametric loudspeaker. PhD thesis. Institute of Acoustics, Chinese
Academy of Sciences; 2011 [in Chinese].
[47] Lawton BW. Damage to human hearing by airborne sound of very high
frequency of ultrasonic frequency. Contract research report 343. HSE Books;
2001.
[48] Guidelines for the safe use of ultrasound: part II – industrial and commercial
applications, safety code 24, Minister of Supply and Services Canada; 1991.
[49] Howard CQ, Hansen CH, Zander AC. A review of current ultrasound exposure
limits. J Occup Health Safety Aust New Zeal 2005;21:253–7.
[50] Lee S, Katsuura T, Towatari-Umeno T, Shimomura Y. The effects of parametric
speaker sound on salivary hormones and a subjective evaluation. Act Nerv
Super Rediviva 2010;52:119–24.